haw-hamburg.de
Mobility Optimization in IPv6 Networks
o Mobility Paradigm
o Review: Mobile IPv6
o Handover Acceleration
o Predictive versus Reactive
o Comparison: SIP Mobility
schmidt@informatik.
haw-hamburg.de
VCoIP in Praxis
haw-hamburg.de
schmidt@informatik.
haw-hamburg.de
VCoIP in Praxis
haw-hamburg.de
IP Mobility Approaches
o Application: SIP Handover
- SIP-server as application specific home agent
o Transport: Mobile SCTP
- Stateful transport handover (doubly bound)
o Multicast-based IP Mobility Support
- Mobile with personal multicast address
o Mobile IPv6
- Stateless, transport transparent handover
schmidt@informatik.
haw-hamburg.de
MIPv6 Release – Mobility on the Rise?
haw-hamburg.de
What may we expect?
o Devices using Home Address while away
o ‘Workspaces’ roaming between local subnets +Improvements on handover performance +3G Mobiles operating IP
+ …
o VoIP/VCoIP conferencing: real-time mobility
o Group communication by Mobile Multicast
Thomas Schmidt schmidt@informatik.
haw-hamburg.de
3GPP/UMTS Release 5 Referenz-Modell
Gf Gi
Iu
Mr Gi Ms
Gi Gc Gr
GGSN EIR
MGCF R-SGW
MRF
Multimedia IP Networks Applications &
Services *)
Mm Mw
Legacy mobile signaling Network
Mc Cx
R Um
TE MT BSS/GRAN
Mh
CSCF
CSCF
Mg
T-SGW *) HSS *)
SCP
CAP
Gi
R Uu
MGW Gn
Signalling and Data Transfer Interface Signalling Interface
TE MT UTRAN PSTN/
Legacy/External
T-SGW *)
HSS *) Applications
& Services *)
GMSC server
*) those elements are duplicated for figure layout purpose only, they belong to the same logical element in the reference model Mc Mc
D
C MGW
Nb
Nc Iu1
Iu
2
R-SGW *) Mh
MSC server SGSN
MS Circuit Switch Access Network
GPRS Access Network
IM Domain
CS Domain PS Domain
Iu A
CAP CAP Alternative
Access Network
Gb
IM Domain is now a sub-set of the PS Domain
UMTS Release 5 requires IPv6 !
haw-hamburg.de
IP Mobility ?
Problem:
Preserve Upper Layer (L 4+) Communication when changing IP Subnets Key Aspects:
- Mobile Node‘s (MN) global adressability (fixed Home Address) - Mobile Node‘s local adressability (changing Care of Address) - Keeping partners informed (updating Correspondent Nodes) - Enabling efficient communication (shortcuts)
Approaches:
Mobile IPv4: IP Mobility Support for IPv4 (RFC 3344)
Mobile IPv6: Mobility Support in IPv6 (RFC 3775/3776)
schmidt@informatik.
haw-hamburg.de
Mobile IP
o IPv4‘s Design Stationary (Routing-Updates Slow)
o Implementation of Mobile Services: Tunneling via Home Agent o IPv6 Potential:
- Several Addresses (2 for Mobile Node, many for Mobile Networks) - Flexible Architecture - no dedicated Access-Services (Agents, DHCP)
Internet
Mobile Node Home
Agent Access Router
haw-hamburg.de
Kommunikations- partner
home agent
foreign agent Mobiler Host
Heimat des mobilen Host
Mobile IPv4
schmidt@informatik.
haw-hamburg.de
Kommunikations- partner
home agent
Mobiler Host
Heimat des mobilen Host
Mobile IPv6
haw-hamburg.de
Mobile IPv6
register
Binding Updates
schmidt@informatik.
haw-hamburg.de
Basic Mobile IPv6
MIPv6 transparantly operates address changes on IP layer by:
o MN‘s stateless configuration of Care of Address in a foreign network and Binding Updates (BUs) with Home Agent (HA) and Correspondent (CNs).
o MN continues to use its original Home Address in a Destination Option Header, thereby hiding different routes to the socket layer.
o CNs continues to use Home Address of the MN, placing current CoA in a Routing Header as Source Route.
o MN, CN & HA keep Binding Cache Tables.
o Home-Agent needed as Address Dispatcher.
haw-hamburg.de
Handover Steps
o Layer 2 Handover
o L3 Movement Discovery o Local Addressing
o Duplicate Address Detection
o Binding Update with Home Agent
o Binding Update with Correspondent Node
schmidt@informatik.
haw-hamburg.de
Handover Security
Binding Udates place a severe security challenge:
MN must prove that it owns claimed IP addresses o BU with HA: IPsec Security Association (strong coupling)
o BU with CN: Return Routablility Procedure (lightweight coupling) to test correctness of MN’s HoA and CoA
- HoTI/HoT: MN(Cookie) → HA → CN (HToken, Cookie) → HA → MN - CoTI/CoT: MN (Cookie) → CN (CToken, Cookie) → MN
- Finally do BU with Hash(HToken, CToken) invertable by CN
o BU improvements: CGA-based (OMIPv6)
haw-hamburg.de
Real-Time Requirements
! Latency ≈< 100 ms
! Jitter ≈< 50 ms
! Packet loss ≈< 1 %
! Interruption: 100 ms ≈ 1 spoken syllable
→ 100 ms are critical bound
Thomas Schmidt schmidt@informatik.
haw-hamburg.de
Local Handover Measurements:
Empirical Results
haw-hamburg.de
L2-Trigger & L2 Handover
IP - Reduce
- MAX_RA_DELAY_TIME ≈ 1 – 3 ms
- MAX_RTR_SOLICITATION_DELAY ≈ 1 – 3 ms
802.11b - Schulzrinne et al.:
Selective Scan + Cache
MIPv6 Handover:
Topology Problem
Thomas Schmidt schmidt@informatik.
haw-hamburg.de
o Generally HA and CN are at Significant Distance o Handover Time: ( t X is RTT MN ↔ X)
o Jitter Enhancement:
o Essential: Eliminate HA/CN RTT Dependence
HA CN
local
CN of
BU HA
of BU local
handoff
t t
t
t t
t t
2 2
3 +
+
≈
+ +
= − − − −
CN
CN HA
stationary handoff
t
t t
Jitter
Jitter +
≈
haw-hamburg.de
Fast MIPv6 (RFC 4068)
Thomas Schmidt schmidt@informatik.
haw-hamburg.de
Reactive Handover with Proxies:
Hierarchical MIPv6 (RFC 4140)
haw-hamburg.de
Predictive versus Reactive
Relevant criteria
►Handover performance: packet loss, delay + jitter
►Number of performed handovers
►Number of processed handovers
►Robustness
►Handover Costs
schmidt@informatik.
haw-hamburg.de
o Compare reactive vers.
predictive handover
o Characteristic to problem:
Router distance
o Charac. to predictive HO:
o Charac. to reactive HO:
Simple analytical model:
3
t l
) (
) 2
( t
Ant− t
l3+ t
L2− t
l32
3 L
l
t
t +
Handover Performance
haw-hamburg.de
More detailed …
o Reactive Handover:
o Predictive Handover (successful):
where
schmidt@informatik.
haw-hamburg.de
Packet Loss Function
L2 Delay: 50 ms
Traffic:
CBR at 1 Pkt/10 ms
haw-hamburg.de
Comparative Samples
schmidt@informatik.
haw-hamburg.de
Stochastic Simulation
o Constant bit rate traffic from CN/HA (at 10 ms) o Random perturbations (ξ) at each link
o Parameters:
- Anticipation Time: <x> = * ms, ξ = 30 ms
- L2 Handoff: <x> = 50 ms, ξ = 10 ms
- Local Links: <x> = 2 ms, ξ = 1 ms
haw-hamburg.de
Simulation: Packet Loss
schmidt@informatik.
haw-hamburg.de
Why is Reality Worse?
Analytical Model did not Account for o Geometry
o Link Perturbation
o Limitations in Completing HO Negotiation
haw-hamburg.de
Negotiation
schmidt@informatik.
haw-hamburg.de
Details on Packet Loss
haw-hamburg.de
Number of Handovers
Relevant quantities:
- Cell residence time - Call holding time - AR-to-MAP ratio
Modelling assumptions:
- Cell residence & call holding time exp. distributed
(homogeneous distribution)
schmidt@informatik.
haw-hamburg.de
Expected # of Handovers
[ ] HO k ρ 1 k 1 ρ
E =
2+
Analytical result:
ρ = Call-to-mobility factor
k = AR-to-MAP ratio
haw-hamburg.de
Stochastic Simulation
Models:
Random Waypoint Varying Geometry
Random Direction
Varying Geometry
Varying Speeds
schmidt@informatik.
haw-hamburg.de
Mean Handover Frequencies:
Random Waypoint Model
haw-hamburg.de
Random Direction Model
schmidt@informatik.
haw-hamburg.de
Erroneous Prediction Yields
haw-hamburg.de
Variation of Interference Radii
schmidt@informatik.
haw-hamburg.de
Handovers for Varying
Interference Radii
haw-hamburg.de
Varying Interference Radii
About 50 %
Bad Predictions
schmidt@informatik.
haw-hamburg.de
Sources of Erroneous Prediction
haw-hamburg.de
Robustness
o Topology
FMIPv6 and HMIPv6 both are unaffected by long distance topology (local ‘step size’ only)
o Rapid Movement
FMIPv6: Forwarding will fail for handover intervals below inter-AR signalling period
HMIPv6: Forwarding will function for any handover
frequency, but delays may increase
schmidt@informatik.
haw-hamburg.de
Application Layer:
Performance of SIP Handover
haw-hamburg.de
Performance of SIP Handover (2)
schmidt@informatik.
haw-hamburg.de
Performance of SIP Handover (3)
haw-hamburg.de
